diff --git a/tests/test_metal_kernel_paged.py b/tests/test_metal_kernel_paged.py
index a1d77b6..4f39b49 100644
--- a/tests/test_metal_kernel_paged.py
+++ b/tests/test_metal_kernel_paged.py
@@ -169,9 +169,6 @@ def qwen3_model():
 
 class TestMetalKernelPagedVsStandard:
     @pytest.mark.slow
-    @pytest.mark.xfail(
-        reason="Metal paged-attention parity mismatch vs standard path (see #119)"
-    )
     def test_greedy_output_matches(self, qwen3_model):
         """Metal kernel paged attention greedy decode must match standard path."""
         model, tokenizer = qwen3_model